Predicting earnings per share using feature-engineered extreme gradient boosting models and constructing alpha trading strategies

This study explores the effectiveness of Extreme Gradient Boosting (XGBoost) models in predicting a stock's future Earnings Per Share (EPS). It utilizes preprocessed technical, fundamental, and analyst forecast data, along with feature engineering techniques to construct indicators. The study a...

Full description

Saved in:
Bibliographic Details
Published in:International journal of information technology (Singapore. Online) 2023-12, Vol.15 (8), p.3999-4012
Main Authors: Singh, Gargi, Thanaya, Indra
Format: Article
Language:English
Subjects:
Artificial Intelligence
Computer Imaging
Computer Science
Image Processing and Computer Vision
Machine Learning
Original Research
Pattern Recognition and Graphics
Software Engineering
Vision
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study explores the effectiveness of Extreme Gradient Boosting (XGBoost) models in predicting a stock's future Earnings Per Share (EPS). It utilizes preprocessed technical, fundamental, and analyst forecast data, along with feature engineering techniques to construct indicators. The study achieves good results in EPS prediction using these feature-engineered datasets and varied hyperparameter settings. The predicted EPS aligns with the actual EPS, demonstrating the model's efficacy. The study proposes using the predicted EPS for developing trading strategies and demonstrates the model's robustness in the Financial and Technological sectors. It also investigates the feature dependency of various technical and fundamental parameters for estimating EPS, highlighting the importance of feature selection to enhance prediction accuracy. Additionally, the study employs these predictions to build alpha trading strategies and portfolios, resulting in anomalous returns with an average Information Ratio of 6.14. The research suggests expanding the use of XGBoost models to anticipate additional fundamental parameters, enabling the construction of more robust portfolios and alpha trading approaches.
ISSN:2511-2104
2511-2112
DOI:10.1007/s41870-023-01450-0